AI and jobs: what the studies really say,
without the headlines.
Every week, a new figure announces the end of millions of jobs, or the opposite. We read the studies in full for you, and we tell you simply what they show.
Six ideas to remember before your next committee meeting.
AI changes tasks, not entire jobs
According to Anthropic (2025), AI is used for at least a quarter of tasks in 36% of jobs, but for three quarters in only 4%. Stanford and UCL (London) point in the same direction.
Teams accept AI, on their own terms
According to Stanford (2026, 1,500 US employees), respondents are in favor of automating 46% of their tasks, first of all to free up time for higher-value work.
Being able to use AI doesn't mean using it
In 2024, 12% of employed people in Europe used an AI that writes or summarizes at work, around 9% in France. Adoption grows when teams are involved in decisions (UCL, 2026).
AI isn't yet reliable on real work
In May 2026, on real professional tasks, the best AI able to act on its own scored 45.9%. The study has not yet been reviewed by other research teams.
The obstacle is human and organizational
According to a preliminary MIT report (2025), the main obstacle is teams' reluctance to adopt new tools.
It's an HR issue, and now a regulatory one
Since February 2025, the EU AI Act has required companies to train their teams on the AI they use. From December 2, 2027, it will also govern so-called “high-risk” HR tools.
Look at tasks, not job titles.
That's where AI acts, and it's the principle behind the Skillberg method. What AI can do doesn't tell you what to hand over to it: your teams' willingness and its real reliability matter just as much. See getting AI adopted and what's at stake for you.
These studies, applied to your company's jobs.
We match each of your tasks to the ones measured by these studies. Let's talk for 30 minutes, and see what a pilot would reveal at your company.
30 minutes on video with Skillberg, no commitment. No personal data required.
- Handa et al., Anthropic Economic Index, 2025 (arXiv:2503.04761)
- Shao et al., Future of Work with AI Agents, Stanford, v3 February 2026 (arXiv:2506.06576)
- Henseke, From Exposure to Adoption, UCL, April 2026 (arXiv:2604.18849)
- JobBench, University of Washington et al., May 2026 (arXiv:2605.26329)
- MIT NANDA, The GenAI Divide: State of AI in Business 2025, July 2025 (preliminary report)
- Regulation (EU) 2024/1689 on AI; postponed to December 2, 2027 by the “Digital Omnibus” regulation